TECHNICAL EFFICIENCY IN MANUFACTURING

INDUSTRIES IN MALAWI USING DETERMINISTIC

PRODUCTION FRONTIER

WC/05/98

Ephraim Wadonda ChirwaUniversity of Malawi and Wadonda Consult

University of MalawiChancellor College, Department of Economics

P.O. Box 280, Zomba, Malawi

Tel: (265) 522 222 Fax: (265) 523 021

Working Paper No. WC/05/98

September 1998

TECHNICAL EFFICIENCY IN MANUFACTURING

INDUSTRIES IN MALAWI USING DETERMINISTIC

PRODUCTION FRONTIER

Ephraim Wadonda Chirwa*

University of MalawiChancellor College, Department of Economics

P.O. Box 280, Zomba, Malawi

Correspondence Address

School of Economics and Social StudiesUniversity of East Anglia

Norwich NR4 7TJ, United KingdomEmail: E.Chirwa@uea.ac.uk

(Revised)

February 1999

I acknowledge valuable comments from two anonymous referees. I am responsible any remaining errors.

See Chirwa (1994) for a detailed analysis of the domestic market and market structure1

and prospects for industrial sector opportunities.

1

TECHNICAL EFFICIENCY IN MANUFACTURING INDUSTRIES INMALAWI USING DETERMINISTIC PRODUCTION FRONTIER

Abstract

This paper uses the census of production data for large establishments in selected manufacturing industries in Malawi

to estimate the level of technical efficiency between 1984 and 1988 using the deterministic production frontier

approach. The study uses panel data for seven manufacturing sectors: tea; tobacco; wearing apparel; printing and

publishing; soaps, perfumes and cosmetics; plastic products and fabricated metal products. We find that the mean

overall technical efficiency ranges from 38 percent in the printing and publishing industry to 87 percent in fabricated

metal products. However, the minimum technical efficiency scores range from as low as 16 percent in the tea

industry to 55 percent in plastic products at firm level. The predicted firm level efficiencies are explained by firm

specific and industry characteristics. The analysis reveals that the market share of the firm is positively associated

with technical efficiency while monopoly power is negatively associated with technical efficiency.

1. Introduction

The manufacturing sector accounts for 12 percent of gross domestic product in Malawi. The

bulk of industrial production is concentrated in five sub-sectors: food processing including sugar

and tea; beverages; tobacco; textiles, netting and blankets; and pharmaceuticals, paints, soaps and

cooking oils. These sectors accounted for about 71 percent of gross output and 78 percent of

total employment in the manufacturing sector. Most of these five sub-sectors have strong1

backward linkages with the agricultural sector as a source of raw materials but very few industrial

linkages within the manufacturing sector. Overall the manufacturing sector in Malawi imports

about two-thirds of their raw materials and sell their products to the final consumer (World Bank,

1989). The manufacturing sector is also highly concentrated and has enjoyed protection of

various forms from several interventionist policies pursued by the government.

The structural adjustment programs that the Malawi government has been pursuing since 1981

have led to liberalisation of output and input markets and have opened the economy to

international competition. These changes in output and input markets and the competitiveness

of economic activities affect the environment in which firms operate and alters incentives for

profit maximisation or cost minimisation. Arguably, since liberalisation of markets alters the

See Kaluwa (1992) and Chirwa (1994, 1998) for cases of potential conflicts among2

objectives.

2

incentive structure and the competitive environment, we would expect such changes to have

influence on the input-output combination of manufacturing activities. Therefore, studying the

efficiency of firms within the adjustment period over time is important in order to determine

wether the incentives created by market reforms lead to improvements in the use of resources.

Previous studies on the performance of firms in the Malawian manufacturing sector focus on

profitability (Kaluwa, 1986; Kaluwa and Reid, 1991), total factor productivity (Mulaga and

Weiss, 1996; Mulaga, 1995; World Bank, 1989) and export performance (Chirwa, 1998).

Technical efficiency is a form of productive efficiency and is concerned with the maximisation

of output for a given set of resource inputs. Productive efficiency reflects the combination which

minimizes the cost of producing a given level of output or equivalently the combination of inputs

that for a given monetary outlay maximizes the level of production. We estimate technical

efficiency using data envelopment analysis (DEA), a non-parametric and deterministic approach

that uses the ‘best practice’ production function measures proposed by Farrell (1957). We

organize the rest of paper as follows. Section 2 presents a brief overview of government policy

towards the manufacturing sector in Malawi. In Section 3, we present the data envelopment

analysis (DEA) technique of estimating technical efficiency. Section 4 presents a description of

data and the methods of estimating technical efficiency. In Section 5, we present and discuss

empirical results on the technical efficiency of firms and sources of technical efficiency. Finally,

Section 6 provides concluding remarks for the paper.

2. Government Policy Towards the Manufacturing Sector in Malawi

Government policies towards industry from independence in 1964 until 1991 have been more

protective and often conflicting with each other and economic objectives. In the first two2

decades of independence the emphasis had been on the development of import substitution

manufacturing industries. This resulted in some protectionist policies through tariff and non-

tariff barriers and a tight regulatory regime for new entrants. Effectively, government

intervention in the manufacturing sector amounted to what we can characterize as anti-

Kirkpatrick et al. (1984) defines an effective competition policy as legislative or3

constitutional elements that essentially control the concentration of economic power, promotecompetition, encourage innovation, create conditions favourable to employment, strive to controlinflation, encourage balanced economic development and promote social welfare.

3

competition policy. We can loosely divide the policy regime into three periods. The first period3

is a protective regime that covered the period between independence and the year before

structural adjustment reforms (1964 - 1980). The second period is the intermediate or adjusting

period in which the government pursued a phased approach to industrial sector reforms between

1981 and 1994. The third period is the post-adjustment era or less protective regime since 1995.

This study focuses on part of the intermediate or adjusting period, particularly the 1984-1988

period.

Industrial policy in the protective regime was characterized by price controls that covered more

than forty-two industrial products, regulation of entry into manufacturing activities through the

licensing procedure that gave incumbent firm strategic advantages, provision for exclusive

monopoly rights to large enterprises with the potential for the exploitation of economies of scale,

presence of institutional barriers to entry and direct state investment in the manufacturing sector.

Kaluwa (1986, 1992), Khan et al. (1992) and Chirwa (1994, 1998) provide extensive discussions

of these issues. Kaluwa (1986) notes that these price controls had little incentives to efficiency

(cost minimisation) and created equity problems and limited entry by restricting profitability that

is the main source of investment financing. Direct state investment enhanced the protection of

markets since most operated as statutory monopolies or oligopolies. Other protective policies

included the allocation of foreign exchange and high tariffs. These policies contributed to the

emergence of a highly monopolistic and oligopolistic market structure in Malawi manufacturing

enterprises.

The intermediate or adjusting regime involves a phased reform of some policies that the

government pursued in the protective regime. The industrial sector reforms were part of the

overall economic reform program sponsored by the World Bank and International Monetary Fund

(IMF) in a series of Structural Adjustment Programs (SAPs). Under structural adjustment

programs various policy changes took place to correct the market distortions. Since 1981, the

See Chirwa and Chilowa (1997) for the sectoral policy actions under each program and4

their effects on trade and industry development, and labour markets.

4

government has tapped seven structural adjustment loans (SALs) and sectoral credits from the

IMF and World Bank to support the implementation of a series of structural adjustment policies.4

Some structural adjustment measures directly aimed at stimulating growth in the manufacturing

sector and are presented in Table 1. Policies towards the industrial sector clearly dominated the

objectives of the first three programs, SAL I - SAL III. The program aimed at diversifying the

export base, encouraging efficient import substitution and ensuring appropriate price and income

policy. The policy actions taken to achieve these objectives centred on devaluation, agricultural

prices and interest rate adjustments, reduction of products under price controls, privatization of

state-owned enterprises.

[Table 1 goes here]

The fourth program, the Industrial and Trade Policy Adjustment Credit (ITPAC), was a special

sectoral credit facility aimed at improving the policy environment for the manufacturing sector

to increase the efficiency of resource use including imports and to expand exports. The policy

of industrial price deregulation continued, and only the prices of fuel and motor vehicle parts

were under price control. The government abolished exclusive monopoly rights in 1988,

reduced the scope of export licensing and liberalized foreign exchange allocation.

The programs that followed ITPAC namely, Agricultural Sector Adjustment Credit (ASAC),

Entrepreneurship Development and Drought Recovery Program (EDDRP) and Fiscal

Restructuring and Deregulation Program (FRDP) reinforced the earlier adjustment policies. For

instance, the government deregulated entry in manufacturing activities in 1992 and reduced

tariffs with the consolidated tariff rates limited to a maximum of 75 percent.

It is evident from the above analysis that the industrial sector policies in particular and the

economic policies overall have attempted to remove price distortions and adjustments to a market

system are almost complete. The policy changes alter the incentive structure of the firms and

See also Charnes et al. (1981) and Bjurek et al. (1990).5

Fare et al. (1983, 1985, 1994) generate six measures of efficiency by relaxing6

assumptions and use of input price data. These measures include overall productive efficiency, allocative(or factor price) efficiency, overall technical efficiency, pure technical efficiency, congestion efficiencyand scale efficiency. A review of some developments in DEA is found in Seiford and Thrall (1990).

See Fare et al. (1994) for the distinctions and linear programs for estimation of efficiency7

(inefficiency).

5

their external environment and influence the input-output mix in production processes. We

expect technical efficiency to improve in the manufacturing sectors during the adjustment period,

for instance, due to increases in domestic and import competition, liberalisation of input and

output prices and increases in trade openness.

3. Data Envelopment Analysis

Data Envelopment Analysis (DEA) is a deterministic and nonparametric approach to efficiency

measurement that has mostly been used in operational research and management science,

following Charnes et al. (1978). This approach is an alternative measure of Farrell's original

approach of computing the efficiency frontier as a convex hull in the input coefficient space by

assuming strong disposability of inputs and constant returns to scale. DEA is a linear5

programming approach for measuring the technical efficiency of a multiple input-multiple output

individual decision making unit (DMU) that does not require any prior assumptions about the

form of the production function. Fare and Lovell (1978), in related work, also provide an

alternative measure of technical efficiency based on the axiomatic treatment of the production

function. The DEA approach developed by Fare and others as presented in Byrnes et al. (1984)

and Fare et al. (1985, 1994) allow further decomposition of technical efficiency into several

mutually exclusive and exhaustive components, and where prices for inputs are available overall

productive efficiency and allocative (factor price) efficiency are estimated by linear programs.6

The measures of technical efficiency are based on output and input sets. Output-oriented7

measures of efficiency determine the extent to which output could be increased given inputs. On

the other hand, input-oriented measures of efficiency identify the extent which inputs could be

6

proportionally reduced to produce a given quantity of output. We follow the Fare tradition and

focus on input-oriented measures of technical efficiency.

3.1 Constant Returns to Scale Model

The constant returns to scale model assumes strong disposability of inputs (S) and constant

returns to scale (C). Strong or free disposability refers to the ability to stockpile or dispose of

unwanted commodities. The linear program minimizes 2 which determines the amount by which

observed inputs can be proportionally decreased if they are utilized efficiently. We solve the

following linear program:

(1)

s.t. i = 1,...,s

r = 1,...,m

where is the overall technical efficiency of the firm , 2 is the measure of

technical efficiency, denotes output r (r = 1, ...., s) for the j th firm, denotes input i (i =

1, ....,m) and are the weights used to construct hypothetical firms on the frontier. The

relative efficiency here captures the percentage by which observed inputs can be proportionally

decreased, given the output, if firms used them efficiently. This is equivalent to measuring the

ratio of actual output to potential/efficient (frontier) output for output-oriented measures.

We illustrate the DEA technique in Figure 1 with one output and one input and four firms A, B,

C and D. Under constant returns to scale and strong disposability, the frontier technology

constructed by the observations is represented by the ray from the origin through point B. In the

7

input-oriented measure of efficiency, only B is efficient since it is on the frontier and the input-

saving measure is calculated by . Firm A is inefficient since output at point A can

potentially be produced by a smaller quantity of input rather than and its

efficiency level is calculated as .

[Figure 1 goes here]

3.2 Variable Returns to Scale Model

We relax the assumption of constant returns to scale for estimating overall efficiency to obtain

efficiency under variable returns to scale (V), while maintaining the assumption of strong

disposability of inputs (S) following Banker et al. (1984). The reference technology in Figure

1 is bounded by and the horizontal line to the right of C and the x-axis from to

infinity. This implies that firms A, B and C are fully efficient, while the efficiency of D is

measured by , which is less than one and below the frontier. Imposing a further restriction

on the weights relaxes the assumption of constant returns to scale. The resulting linear program

derives the input-oriented Weak Efficiency measure under variable returns to scale:

(2)

subject to a further restriction in (1)

We calculate another input-oriented measure of technical efficiency under the assumption of

nonincreasing returns to scale (N) and strong disposability of inputs (S) (Fare et al., 1994; Bjurek

et al., 1990). We solve the linear program in (1) to obtain a Weak Efficiency measure under

nonincreasing returns to scale by imposing a further constraint. Thus,

(3)

8

subject to an additional constraint in equation (1)

The technology for in Figure 1 is bounded by the x-axis and OBC and the

horizontal line to the right of C. Thus, point A which is efficient under variable returns to scale

and strong disposability, is inefficient under nonincreasing returns to scale and strong

disposability of inputs. Points B and C are technically efficient both under variable returns to

scale and nonincreasing returns to scale.

3.3. Scale Efficiency

Scale efficiency captures departure of a firm from optimal scale. The input-oriented scale

efficiency measure is given as:

j = 1,2, ...., J (4)

Thus, firm j is input scale efficient if , or if it is equally technically efficient

relative to the (C,S) and (V,S) input set. Scale efficiency that measures input loss due to

operating at an inefficient scale. In Figure 1, only point B is scale efficient. While points A, C,

and D are scale inefficient since they could produce the same output with the less inputs if they

operated on an efficient scale (the difference between nonconstant returns to scale technology and

the constant returns to scale technology).

These four measures of efficiency enable the identification of the types of returns to scale in

production for a particular firm. We have constant returns to scale if =

. This is satisfied at point B. If = �

, then the unit under consideration produces at decreasing returns to scale. The

firm at point C produces at decreasing returns to scale. Finally, production takes place at

increasing returns to scale if � , such as at points A and D.

See Lovell (1993), Charnes et al. (1995) and Tulkens and Eeckaut (1995).8

9

In all the above types of technical efficiency, production is technically efficient if the measure

of efficiency is equal to unity. If we have technical inefficiency, the corresponding measures will

be less than one. The difference between unity and the observed measure yield the percentage

of potential input savings that the firm could make due to the particular type of inefficiency.

3.4 DEA Approaches with Panel Data

The literature suggests at least four approaches of estimating technical efficiency in data

envelopment analysis with panel data. These approaches are the window analysis, the

contemporaneous frontier, sequential frontier and the intertemporal frontier. In the present8

study we use the window analysis and the intertemporal frontier. The former ensures that

technology is never forgotten while later increases the number of observations when the number

of firms in the industry is small.

The window analysis technique involves partial pooling of panel data (k enterprises over T time

periods). The panel is divided into a series of shorter overlapping panels, each having k

enterprises and W time periods. The first pooled panel (window) consists of k enterprises and

time periods {1,...,W}, the second consists of k enterprises and time periods {2,...,W+1}, and so

on until the final window with k enterprises and time periods {T-W+1,...,T}. Thus, the window

is shifted forward one period each time. The widths of each window (W) are equal and arbitrary.

Each enterprise gets one efficiency score in the first period with a comparison set consisting all

enterprises in time periods {1,...,W}; two efficiency scores in the second period, with the first

comparison set consisting of all enterprises in time periods {1,...,W} and the second comparison

set consisting of all enterprises in time periods {2,...,W+1}; and so on. The window analysis

technique increases the degrees of freedom and reveals the trends in efficiency scores.

Tulkens and Eeckaut (1995) propose the estimating efficiency from the ‘intertemporal’ frontier

in which DEA is applied on the pooled cross-section and time-series sample. This assumes that

the reference technology is time invariant. The pooling of cross-section and time-series data has

See Zhang and Bartels (1998) on the implications for sample size in comparing9

efficiency across activities.

The National Statistical Office only publishes data for large establishments with a sales10

turnover of K100,000 at 1978 prices.

10

the advantage of increases the number of observations especially in cases where the number of

firms in the industry is small.

The major advantage of data envelopment analysis is that it does not require a specified

production function or the weights for different inputs and outputs used (Seiford and Thrall,

1990). However, one disadvantage of the DEA approach is that it is sensitive to the number of

observations, extreme observations and measurement errors. The DEA approach is used9

extensively in measuring efficiency in various economic activities. Coelli (1995) provides a

review of some studies in agriculture economics that have used DEA to estimate technical

efficiency and Seiford (1996) gives a bibliography of theoretical and empirical work in data

envelopment analysis. Other recent studies that apply the DEA technique include Cowie and

Riddington (1996), Arnade (1998), Burki and Terrell (1998).

4. Data and Estimation

The study uses data from the census of production collected by the National Statistical Office

(NSO) in the Annual Economic Survey. We estimate technical efficiency for large

establishments in selected seven four-digit International Standard Industrial Classification (ISIC)

industries between 1984 and 1988. The data set that includes medium and small scale enterprises

does not contain values for capital and raw materials. Only industries that have at least four10

firms were selected in this analysis. We use the concept of a firm producing one output and

several inputs to compute input-oriented measures of technical efficiency.

DEA is performed using the window analysis technique and the intertemporal frontier. In the

former we use the two-year windows for each respective sector leading to four windows in the

five-year period. In the latter we pool cross-section and time-series and estimate one frontier for

each sector. Output is measured by sales at 1980 prices. Inputs include number of workers

11

(labour), real capital (capital) estimated as real net book value of assets at end-of-year valued at

1980 prices, and real value of raw materials (materials) at 1980 prices. The technical efficiency

scores are obtained from the DEAP 2.1, a data envelopment analysis computer program written

by Coelli (1996). Table 2 presents descriptive statistics for the variables used in the estimation

of production frontiers for various industries for the pooled sample. The statistics for the pooled

sample mask the ‘within the period’ variations. However, the data show wide variations across

observations. The standard deviation for most industry groups for output and inputs being

greater than the average and the range suggests a wide gap between the largest and smallest firm

based on the value of sales.

[Table 2 goes here]

5. Empirical Results

5.1 Efficiency Levels in Malawi Manufacturing Industries

Estimates of technical efficiency using panel data are computed using data envelopment analysis

from both the window analysis technique and the intertemporal frontier approach. We report

two types of efficiency - overall technical efficiency and scale efficiency. Table 3 presents the

frequency distribution of technical efficiency scores by industry sector for the whole sample

period. Overall, the mean technical efficiency scores obtained from the window analysis are

higher than those obtained in the intertemporal frontier approach. We discuss results mainly

based on the intertemporal frontier approach. In the tea industry, the mean technical efficiency

is 41 percent and is as low as 16 percent. On average, firms in the tea industry could save 59

percent of inputs to produce existing output levels efficiently. The frequency distribution shows

that 75.1 percent of observations have efficiency scores of less than 50 percent. Only 5 percent

of the observations are technically efficient, thus four technically efficient observations from four

firms.

With respect to the tobacco industry, the mean technical efficiency is 56 percent and is as low

as 19 percent. Half of the observations have efficiency scores of less than 50 percent and only

15 percent of the observations are technically efficient (three technically efficient observations

12

from two firms). The wearing apparel industry records a mean technical efficiency score of 69

percent and the minimum score in the sector is 31 percent. More than 80 percent of the

observations have efficiency scores between 60 percent and 100 percent. Twenty percent of the

observations, from six firms, are technically efficient.

[Table 3 goes here]

The technical efficiency scores in the printing and publishing industry are as low as 17 percent

with a mean of 38 percent. About 75 percent of the observations have efficiency scores below

40 percent. Only 10 percent of the observations from two firms are technically efficient. With

respect to the soaps and cosmetics industry, we observe a high mean efficiency of 79 percent with

the lowest efficiency score of 38 percent. Of the total observations in the industry, 85 percent

have efficiency scores between 60 percent and 100 percent. The soaps and cosmetics industry

has four technically efficient observations from two firms.

The mean technical efficiency in the plastic products industry is 82 percent with a minimum

efficiency score of 55 percent. The efficiency scores from the window analysis are similar. Most

of the observations are close to the frontier, with 92 percent of efficiency scores falling between

60 percent and 100 percent. The industry also records six technically efficient observations but

from only two enterprises. Similarly, most observations in the fabricated metal products industry

are close to their frontier, with a mean efficiency of 87 percent and a minimum efficiency score

of 49 percent. Most observations are technically efficient (40 percent), thus eight on the frontier

spanning four firms. Thus, at least each firm is technically efficient for the pooled sample in the

sector. We obtain similar results from the window analysis.

Overall, if we take technical efficiency scores above 80 percent, we find that proportionately

most observations in the fabricated metal products industry (65 percent) are more efficient

compared with 56 percent in plastic products, 55 percent in soaps and cosmetics, 35 percent in

wearing apparel, 30 percent in tobacco industry, 10 percent in printing and publishing and 7.6

percent in the tea industry. This ordering is also apparent from the frequency distribution based

on the window analysis efficiency scores.

13

Table 4 shows the mean annual technical efficiency levels by industry sector based on both the

window analysis and intertemporal approach. We focus on the results from the intertemporal

frontier approach. The mean overall technical efficiency ranges from 36 percent in 1986 to 49

percent in 1984 while the mean scale efficiency is stable ranging from 75 percent in 1988 to 78

percent in 1985 for the tea industry. In the tobacco industry, the mean overall technical

efficiency ranges from 46 percent to 70 percent and the trend is a declining one over time.

However, mean scale efficiency ranges from 70 percent in 1988 to 90 percent in 1985. The mean

annual overall technical efficiencies in the wearing apparel industry are also stable but tend to

increase over time. The trend in mean annual scale efficiency is similar ranging from 83 percent

in 1984 to 90 percent in 1988.

[Table 4 goes here]

The mean annual overall technical efficiency in the printing and publishing industry shows a

clear declining trend over time from 42 percent in 1984 to 32 percent in 1988. Similarly, the

mean annual scale efficiency ranges from 63 percent in 1985 to 45 percent in 1988. With respect

to the soaps and cosmetics industry, the mean annual overall technical efficiency ranges from 71

percent in 1984 to 87 percent in 1987, generally showing an increasing trend over time. Annual

overall technical efficiencies are stable in the plastic products industry, 83 percent in the first four

years and dropping to 77 percent in the fifth year, however scale efficiencies show a declining

trend over the period of analysis. The trend in both overall and scale efficiency in fabricated

metal products is mixed, but the efficiency levels are above 70 percent and 80 percent,

respectively.

The annualized technical efficiency scores for respective industry frontiers show that mean scale

efficiencies are much higher than overall technical efficiencies. Using the intertemporal frontier

approach we observe that most firms operate at increasing returns to scale. Table 5 presents a

summary of returns to scale by industry sector based on the intertemporal frontiers for respective

industries. Overall, about 56 percent of the observations exhibit increasing returns to scale, while

28 percent exhibit decreasing returns and 15 percent exhibit constant returns to scale. Tea and

wearing apparel industries dominate cases that exhibit increasing returns to scale (44.9% and

14.2%, respectively), plastic products (11.8%), soaps and cosmetics (10.2%).

The available data does not have information of some variables such as the number of11

plants per enterprise, advertising expenses, R and D expenditures and multinationality of enterprises.

14

[Table 5 goes here]

About 71 percent of observations in the tea industry, 65 percent in soaps and cosmetics, 60

percent in plastic products, 50 percent in tobacco, 45 percent in wearing apparel industry and 40

percent in printing and publishing industry exhibit increasing returns to scale. About 24 percent

of observations in tea industry, 35 percent in tobacco, 43 percent in wearing apparel, 50 percent

in printing and publishing, 30 percent in fabricated metal products exhibit decreasing returns to

scale. Thus, in many industries inefficiency is partially due to firms operating at sub-optimal

scale.

5.2 Determinants of Technical Efficiency

5.2.1 Specification of the Model

We also investigate the sources of the predicted inefficiencies using firm and industry attributes

following Caves (1992a, 1992b), Mayes et al. (1994), Brada et al. (1997) and Burki and Terrell

(1998). We use the efficiency scores and relate them to firm and industry characteristics and

policy variables. We estimate the following relationship using pannel data regression:

(6)

where for firm i and industry j at time t, TE is the measure of technical efficiency obtained from

DEA (overall technical and scale efficiency), MS is the market share of the firm, HHI is the

measure of domestic market power, IMPS is the measure of the import competition, EXPS is the

measure of export-orientation, GROW is the market growth for the industry, TARF are nominal

tariffs for imports of each industry and IND are industry dummies. Table 6 presents the

variables included in the regression model. Although literature suggests several factors that

influence firm efficiency, we focus on testing a few hypotheses. 11

The effective protection rate would be more appropriate to use in order to capture the12

real extent of a protective trade regime.

15

[Table 6 goes here]

The market share of the firm (MS) is the firm characteristic variable under the hypothesis that

large market shares reflect the relative efficiency of firms following Demsetz’s (1973) efficient

market hypothesis. The market share also reflects the relative size of the firm, and hence the

importance of economies of scale. MS is calculated as the share of the firms output in the

industry. We expect market share to be positively associated with technical efficiency. Due to

endogeneity of market share, since firms get market shares partly because they are efficient, we

use the lagged market share to instrument this variable in the regression analysis.

The structure of the industry plays a critical role in enhancing efficiency as far as competitive

process is concerned. The nature of domestic competition is reflected by the Herfindahl-

Hirschman index (HHI) which is calculated as sum of squared market shares (based on output)

of all firms in the industry. We expect competition to have a positive relationship with the firm’s

efficiency, hence a negative relationship between technical efficiency and the Herfindahl-

Hirschman index. To capture the role of international competition we use import shares (IMPS)

and export shares (EXPS). IMPS is calculated as the ratio of imports of manufactured imports

for the industry to total domestic supply. The inflows of imports exert some competitive pressure

on domestic firms that in turn should encourage domestic firms to operate efficiently. EXPS

represents the ratio of exports to total industry production. The desire for firms to become

competitive in international markets should lead to higher technical efficiency. We expect EXPS

to have a positive relationship with technical efficiency. GROW is the annual growth of the

industry in real industrial output, which we include to capture demand effects on technical

efficiency.

The role of trade policy is captured by the level of nominal tariffs (TARF) of imported finished

goods in the relevant industries. This measure captures the effect of trade protection on the12

technical efficiency of domestic firms. We expect a negative relationship between TARF and

technical efficiency. We also include industry specific dummies representing the seven industries

16

included in the study: TEA for tea, TOB for tobacco, WEA for wearing apparel, PRI for printing

and publishing, SOA for soaps, perfumes and cosmetics, PLA for plastic products and FAB for

fabricated metal products.

5.2.2 Empirical Results from Panel Regression Model

The regression results of the technical efficiency models using two measures of technical

efficiency, based on the window and intertemporal efficiency scores are presented in Table 7.

We carried out the estimations on TSP 4.4 using panel data methods to obtain the fixed effects

and random effects models. The Hausman specification test is reported for the suitability of the

random effects model and is only in favour of the random effect model in the overall technical

efficiency model based on intertemporal frontier. Discussion of results focuses on our preferred

fixed effect models in equations (1), (5) and (7) and the random effect model in equation (4).

The explanatory power of these models ranges from 45 percent to 69 percent.

[Table 7 goes here]

The results show that the lagged market share of the firm (MS) is positively associated with

technical efficiency in all preferred models. The coefficients are significantly significant at 1

percent level in equations (1), (2) and (7) and at 5 percent level in equation (5). The significance

of the market share variable provides evidence in favour of the efficient market hypothesis that

firms that secure larger market shares are more efficient. These results are also in favour of the

argument that large firms, with economies of scale, are more efficient than small firms.

The level of domestic industrial concentration as indicated by HHI, is negatively associated with

technical efficiency in all the preferred specifications, but the coefficients are only statistically

significant at 10 percent level in equation (1) and at 5 percent level in equation (5). This

relationship reflects the adverse impact of domestic monopoly power on firms’ technical

efficiency and renders some support for the competition hypothesis.

The demand variable (GROW) and the trade policy variable (TARF) are sensitive to the measure

of technical efficiency and are weakly significant and negatively related to technical efficiency

17

in equation (5) at 10 percent level for the former and in equation (4) at 5 percent level for the

later. Firms in the tea industry (TEA) have consistently lower efficiency levels in all the models

and the coefficients are statistically significant. Other industry dummies are weakly significant.

6. Conclusions

This study has estimated the level of technical efficiency in Malawian manufacturing industries

using the deterministic (nonparametric) data envelopment analysis approach. We have used panel

data for seven manufacturing industries over the period 1984 to 1988. The results indicate high

levels of technical inefficiency in most sectors. On average, for firms in Malawian

manufacturing to produce existing levels of outputs, they can save inputs by 59 percent in the tea

industry, 44 percent in the tobacco industry, 41 percent in the wearing apparel industry, 62

percent in the printing and publishing industry, 21 percent in the soaps and cosmetics industry,

18 percent in the plastic products industry and 13 percent in the fabricated metal products

industry.

The results also show that 15 percent of observations exhibit constant returns to scale, while 28

percent and 56 percent exhibit decreasing and increasing returns to scale, respectively. This

shows that inefficiency is partly a result of operating at sub-optimal scale of production. The

average annual efficiency scores for most sectors also show a declining trend based on the

intertemporal frontier approach although this varies from industry to industry. Decreasing trends

are observed in tea, tobacco and printing and publishing industries, while clearly visible

increasing trends are observed in wearing apparel and soaps and cosmetics. Average efficiency

scores are more stable in plastic products and fabricated metal products industries. The trend in

the efficiency of firms in the manufacturing sector during this adjustment period is rather mixed,

with a positive effect in other sectors and negative effects on other sectors.

The second empirical part of the study attempts to identify sources of technical efficiency in the

selected manufacturing industries using a censored regression model. The results show that the

market share of the firm is positively associated with the level of technical efficiency reflecting

the importance of economies of scale. Domestic monopoly power is negatively associated with

18

the level of technical efficiency, suggesting that polices that aim at encouraging domestic

competition are likely to have postive effects on the efficiency of manufacturing enterprises.

References

Arnade, C. (1998) ‘Using a Programming Approach to Measure International AgriculturalEfficiency and Productivity’, Journal of Agricultural Economics, 49 (1), 67-84

Banker, R. D., Charnes, A. and Cooper, W. W. (1984) ‘Some Models for Estimating Technicaland Scale Inefficiencies in Data Envelopment Analysis’, Management Science, 30 (9),1078-1092

Bjurek, H., Hjalmarsson, L. and Forsund, F. R. (1990) ‘Deterministic Parametric andNonparametric Estimation of Efficiency in Service Production’, Journal ofEconometrics, 46, 213-227

Brada, J. C., King, A. E. and Ma, C. Y. (1997) ‘Industrial Economics of the Transition:Determinants of Enterprise Efficiency in Czechoslovakia and Hungary’, OxfordEconomic Papers, 49, 104-127

Burki, A. A. and Terrell, D. (1998) ‘Measuring Production Efficiency of Small Firms inPakistan’, World Development, 26 (1), 155-169

Byrnes, P., Fare, R. and Grosskopf, S. (1984) ‘Measuring Productive Efficiency: An Applicationto Illinois Strip Mines’, Management Science, 30, 671-681

Caves, R. E. (1992a) ‘Determinants of Technical Efficiency in Australia’, in R. E. Caves (ed)Industrial Inefficiency in Six Nations, Cambridge, MA: MIT Press

Caves, R. E. (1992b) Industrial Inefficiency in Six Nations, Cambridge, MA: MIT Press

Charnes, A., Cooper, W. W. and Rhodes, E. (1981) Evaluating Program and ManagerialEfficiency: An Application of Data Envelopment Analysis to Program Follow Through,Management Science, 26 (6), 668-697

Charnes, A., Cooper, W. W. and Rhodes, E. (1978) ‘Measuring the Efficiency of DecisionMaking Units’, Journal of Operational Research, 2, 429-444

Charnes, A., Cooper, W. W., Lewin, A. Y. and Seiford, L. M. (1995) Data Envelopment

Analysis: Theory, Methodology, and Application, Boston: Kluwer AcademicPublishers

19

Chirwa, E. W. (1994) ‘Malawi: Industrial Sector Opportunities - The Domestic Market andMarket Structure’, Department of Economics, University of Malawi (Zomba): Reportprepared for Maxwell Stamp Plc, London

Chirwa, E. W. (1998) ‘Market Structure and the Export Performance of Manufacturing Industriesin Malawi’, Wadonda Consult Working Paper WC/02/98, Wadonda Consult, Zomba

Chirwa, E. W. and Chilowa, W. (1997) Structural Adjustment Programmes in Southern Africa:The Case of Malawi, paper submitted to International Labour Organization, Geneva

Coelli, T. J. (1995) ‘Recent Developments in Frontier Modelling and Efficiency Measurement’,Australian Journal of Agricultural Economics, 39 (3), 219-245

Coelli, T. J. (1996) ‘A Guide to DEAP Version 2.1: A Data Envelopment Analysis (Computer)Program’, CEPA Working Paper 96/07, University of New England, Australia

Cowie, J. and Riddington, G. (1996) ‘Measuring the Efficiency of European Railways’, AppliedEconomics, 28, 1027-1035

Demsetz, H. (1973) ‘Industry Structure, Market Rivalry and Public Policy’, Journal of Law andEconomics, 16 (1), 1-19

Fare, R., Grosskopf, S. and Logan, J. (1983) ‘The Relative Efficiency of Illinois ElectricUtilities’, Resources and Energy, 5, 349-367

Fare, R., Grosskopf, S. and Logan, J. (1985) ‘The Relative Performance of Publicly-Owned andPrivately-Owned Electric Utilities’, Journal of Public Economics, 26, 89-106

Fare, R., Grosskopf, S. and Lovell, C. A. K. (1994) Production Frontiers, Cambridge:Cambridge University Press

Fare, R. and Lovell, C. A. K. (1978) ‘Measuring the Technical Efficiency of Production’,Journal of Economic Theory, 19 (October), 150-162

Farrell, M. J. (1957) ‘The Measurement of Productive Efficiency’, Journal of Royal StatisticalSociety Series A 120, part 3, 253-281

Kaluwa, B. M. (1986) ‘Barriers to Entry, Price Controls and Monopoly Power in MalawianManufacturing’, University of Edinburgh, Department of Economics, Ph.D Thesis

Kaluwa, B. M. (1992) ‘Malawi Industry: Problems, Policies and Performance’, in Mhone G. C.,Malawi at the Cross-roads: A Post-Political Economy, Harare: SAPES

Kaluwa, B. M. and Reid, G. C. (1991) ‘Profitability and Price Flexibility in Manufacturing of aDeveloping Country’, Journal of Industrial Economics, 39, 6, pp 689-700

20

Khan, S., Kaluwa, B. and Spooner, N. (1989) The Impact of Malawi’s Industrial PriceDecontrol Programme, Washington: World Bank

Kirkpatrick, C. H., Lee, N. and Nixson, E. I. (1984) Industrial Structure and Policy in LessDeveloped Countries, London: George Allen and Unwin

Lovell, C. A. K. (1993) ‘Production Frontiers and Productive Efficiency’, in Fried, H. O., Lovell,C. A. K., Schmidt, S. S. (eds.) The Measurement of Productive Efficiency:Techniques and Applications, New York: Oxford University Press

Mayes, D., Harris, C. and Lansbury, M. (1994) Inefficiency in Industry, New York: HarvesterWheatsheaf

Mulaga, G. (1995) Trade Policy and Manufacturing Sector Performance in Malawi, 1970-1991,Bradford: University of Bradford (PhD thesis)

Mulaga, G.and Weiss, J. (1996) Trade Policy and Manufacturing Sector Performance in Malawi,1970-91, World Development, 24 (7), 1267-1278

Seiford, L. M (1996) ‘Data Envelopment Analysis: the Evolution of the State of the Art (1978-1995)’, Journal of Productivity Analysis, 7 (2/3), 99-137

Seiford, L. M. and Thrall, R. A. (1990) ‘Recent Developments in DEA - The MathematicalProgramming Approach to Frontier Analysis’, Journal of Econometrics, 46, 7-38

Tulkens, H. and Eeckaut, P. van (1995) ‘Non-parametric Efficiency Analysis, Progress andRegress Measures for Panel Data: Methodological Aspects’, European Journal ofOperational Research, 80 (3), 474-499

World Bank (1989) Malawi Industrial Sector Memorandum, Washington: World Bank

World Bank (1996) Malawi: Fiscal Restructuring and Deregulation Program, Washington:World Bank

Zhang, Y. and Bartels, R. (1998) The Effect of Sample Size on the Mean Efficiency in DEA withApplication to the Electricity Distribution in Australia, Sweden and New Zealand,Journal of Productivity Analysis, 7 (2-3), 187-204

21

Figure 1 Data Envelopment Analysis Efficiency Measures

22

Table 1 Summary of Industrial Policy Actions under Structural Adjustment Programs

Program/Loan

and Fiscal Years*

Main Policy Actions affecting the Manufacturing Sector

SAL I

1981-83

! Devaluation of Malawi Kwacha by 15% in April 1982, 12% in September

1983 and 3% in January 1984

! Annual increases in smallholder producer prices

! Periodic adjustments of interest rates

SAL II

1984-85

! Devaluation of Malawi Kwacha by 15% in April 1985 and 9.5% in January

1986

! Annual increases in smallholder producer prices

! Industry price decontrol - 41% of controlled products

! Periodic adjustments of interest rates

SAL III

1986-87

! Devaluation of Malawi Kwacha by 10% in August 1986, 20% in February

1987and 15% in January 1988

! Annual increases in smallholder producer prices

! Industrial price decontrol

! Privatization of state-owned enterprises

! Increases in taxes

! Establishment of an export financing facility

! Periodic adjustments of interest rates

ITPAC

1988-89

! Devaluation of Malawi Kwacha by 7% in March 1990

! Industrial price decontrol

! Abolition of exclusive product (monopoly) rights

! Revision of duty drawback system and introduction of surtax credit system

! Partial liberalization of foreign exchange rationing on 65% of imports

! Reduction in the scope of export licensing

ASAC

1990-91

! Annual increases in smallholder producer prices

! Periodic adjustments of interest rates

! Complete liberalization of foreign exchange allocation

EDDRP

1992-95

! Devaluation of Malawi Kwacha by 15% in June 1992, 22% in July 1992

! Floatation of the Malawi Kwacha in February 1994

! Establishment of Malawi Investment Promotion Agency

! Replacement of Industrial Development Act with Industrial Licensing Act

! Review of labour market imperfections and policy option including minimum

wage policy

! Increase in surtax on electricity and telephone services

! Reduction in base surtax rate to 30 percent

! Reduction in tariffs and consolidated tariffs limited to a maximum of 75

percent

! Decreases in corporate and income tax rates

! Elimination of direct bank credit controls and liberalisation of interest rates

FRDP

1996-98

! Implementation of the Export Processing Zones Act

! Establishment of the Malawi Stock Exchange

! Privatization of state owned enterprises

Source: Chirwa and Chilowa (1997) and World Bank (1996).

Notes: * The fiscal year starts in April and ends in March the following year.

23

Table 2 Descriptive Statistics for Output and Inputs by Industry, 1984-88

Industry/(ISIC) Statistic Output Labour Capital Materials Firms

(Obs)

Tea (3123)

Tobacco (3140)

Wearing Apparel excl

Footwear (3220)

Printing and Publishing

(3420)

Soaps, Perfumes and

Cosmetics (3523)

Plastic Products (3560)

Fabricated Metal

Products (3819)

Mean

S. D

Minimum

Maximum

Mean

S.D

Minimum

Maximum

Mean

S.D

Minimum

Maximum

Mean

S.D

Minimum

Maximum

Mean

S.D

Minimum

Maximum

Mean

S.D

Minimum

Maximum

Mean

S.D

Minimum

Maximum

2,981,689

2,275,617

189,804

9,384,339

3,051,874

2,878,014

239,539

8,112,494

780,491

748,594

160,581

3,020,127

2,285,169

2,994,054

176,517

9,255,048

7,217,284

10,755,668

135,287

26,713,917

1,421,126

653,472

468,805

2,633,984

2,190,234

2,422,627

71,589

8,865,572

371

207

90

1,011

1,092

1,462

29

4,759

147

135

12

450

233

252

4

724

161

206

22

591

97

36

44

189

194

152

13

421

139,665

116,304

11,223

451,306

504,203

323,190

88,454

1,113,502

73,737

77,152

12,606

416,995

310,538

403,636

15,290

1,146,262

499,090

796,164

6,947

3,055,805

140,331

79,700

32,329

277,923

206,323

174,648

830

520,100

2,213,193

1,633,581

101,705

6,992,388

1,032,943

771,148

7,864

2,077,709

487,021

491,165

55,412

1,860,127

1,116,028

1,437,778

10,392

4,134,604

3,359,467

5,045,080

48,838

13,980,121

725,475

444,349

117,694

1,675,030

1,193,566

1,491,365

35,392

5,337,567

16

(80)

4

(20)

8

(40)

4

(20)

4

(20)

5

(25)

4

(20)

Note: All values for output, capital and materials are in Malawi Kwacha measured at 1980 prices using the

consumer price index. The exchange rate in 1980: 1 US Dollar = MK 1.25.

24

Table 3 Distribution of Levels of Technical Efficiency by Industry Sector (percent) *

Level of Efficiency

(OTE)

Tea Tobacco WearingApparel

Printing andPublishing

Soaps, Perfumesand Cosmetics

Plastic Products Fabricated MetalProducts

0.1 - < 0.20.2 - < 0.30.3 - < 0.40.4 - < 0.50.5 - < 0.60.6 - < 0.70.7 - < 0.80.8 - < 0.90.9 - < 1.0

1.0

5.031.3 16.322.511.3 3.8 2.5 1.3 1.3 5.0

(1.3)(7.5)(8.8)

(17.5)(16.3)(15.0)

(8.8)(6.3)(7.5)

(11.3)

5.0 5.035.010.010.0 5.0

--

15.015.0

(-)(-)

(15.0)(25.0)(10.0)(10.0)(10.0)

(-)(10.0)(20.0)

- -

10.010.012.515.017.520.0 2.512.5

(-)(-)

(7.5)(2.5)(7.5)

(15.0)(20.0)(17.5)(12.5)(20.0)

10.045.020.0

-10.0

- 5.0

--

10.0

(10.0)(15.0)(20.0)

(-)(-)

(15.0)(5.0)

(10.0)(5.0)

(20.0)

--

5.05.05.0

10.020.025.010.020.0

(-)(-)(-)

(5.0)(-)

(10.0)(10.0)(20.0)(20.0)(35.0)

----

8.020.016.020.012.024.0

(-)(-)(-)(-)(-)

(16.0)(28.0)(12.0)(16.0)(28.0)

---

5.0-

5.025.0 5.020.040.0

(-)(-)(-)(-)(-)(-)

(25.0)(-)

(20.0)(55.0)

CasesMean OTESDMinimumEfficient CasesEfficient Firms

800.410.210.16

44

(80)(0.62)(0.23)(0.19)

(9)(5)

200.560.290.19

32

(20)(0.66)(0.25)(0.30)

(4)(2)

400.690.200.31

54

(40)(0.78)(0.19)(0.35)

(8)(6)

200.380.240.17

22

(20)(0.59)(0.31)(0.17)

(4)(2)

200.790.180.38

42

(20)(0.87)(0.16)(0.40)

(7)(4)

250.820.150.55

62

(25)(0.85)(0.13)(0.61)

(7)(3)

200.870.150.49

84

(20)(0.93)(0.10)(0.74)

(11)(4)

Frequency distributions based on intertemporal frontier (pooled sample) for respective sectors and the figures in parentheses are based on the window*

analysis.

25

Table 4 Summary of DEA Technical Efficiency Indices by Industry Sector, 1984-88

Industry Year

Window Analysis Intertemporal Frontiera b

Overall Technical

Efficiency

(OTE)

Scale

Efficiency

(STE)

Overall Technical

Efficiency

(OTE)

Scale

Efficiency

(STE)

Mean SD Mean SD Mean SD Mean SD

Tea 1984

1985

1986

1987

1988

0.72

0.70

0.70

0.53

0.44

0.21

0.18

0.19

0.19

0.26

0.81

0.82

0.84

0.78

0.74

0.19

0.18

0.18

0.19

0.23

0.49

0.42

0.36

0.37

0.42

0.26

0.16

0.15

0.18

0.25

0.77

0.78

0.76

0.76

0.75

0.21

0.22

0.19

0.20

0.23

Tobacco 1984

1985

1986

1987

1988

0.61

0.70

0.70

0.61

0.66

0.29

0.35

0.24

0.29

0.26

0.77

0.90

0.82

0.83

0.83

0.23

0.14

0.14

0.16

0.16

0.61

0.70

0.57

0.46

0.46

0.29

0.35

0.28

0.32

0.32

0.77

0.90

0.74

0.73

0.70

0.23

0.14

0.25

0.19

0.21

Wearing

Apparel

1984

1985

1986

1987

1988

0.73

0.78

0.76

0.81

0.80

0.22

0.23

0.21

0.14

0.19

0.84

0.86

0.87

0.87

0.84

0.15

0.18

0.17

0.14

0.18

0.67

0.71

0.64

0.68

0.75

0.17

0.22

0.19

0.20

0.25

0.83

0.83

0.85

0.86

0.90

0.14

0.16

0.13

0.12

0.16

Printing and

Publishing

1984

1985

1986

1987

1988

0.42

0.46

0.46

0.71

0.91

0.39

0.36

0.26

0.13

0.12

0.53

0.59

0.59

0.81

0.93

0.32

0.30

0.24

0.15

0.08

0.42

0.46

0.37

0.33

0.32

0.39

0.37

0.23

0.12

0.13

0.54

0.63

0.53

0.52

0.45

0.31

0.29

0.22

0.18

0.08

Soaps,

Perfumes and

Cosmetics

1984

1985

1986

1987

1988

0.90

0.82

0.78

0.91

0.94

0.19

0.16

0.26

0.13

0.05

0.90

0.85

0.81

0.97

0.95

0.20

0.17

0.28

0.04

0.05

0.71

0.72

0.77

0.87

0.85

0.20

0.21

0.27

0.16

0.09

0.81

0.81

0.81

0.94

0.94

0.26

0.23

0.27

0.08

0.06

Plastic

Products

1984

1985

1986

1987

1988

0.86

0.86

0.87

0.86

0.80

0.14

0.14

0.14

0.16

0.14

0.89

0.91

0.93

0.89

0.85

0.15

0.11

0.12

0.17

0.11

0.83

0.83

0.83

0.83

0.77

0.14

0.16

0.16

0.19

0.17

0.91

0.94

0.93

0.89

0.84

0.15

0.11

0.14

0.19

0.11

Fabricated

Metals

1984

1985

1986

1987

1988

0.91

1.00

0.92

0.84

0.99

0.12

0.00

0.12

0.11

0.02

0.99

1.00

0.99

0.94

1.00

0.02

0.00

0.01

0.10

0.00

0.88

1.00

0.84

0.73

0.90

0.13

0.00

0.14

0.20

0.13

0.98

1.00

0.93

0.99

1.00

0.04

0.00

0.13

0.01

0.01

Notes:

Window analysis based on two year windows.a

This is based on the pooled cross-section and time-series frontier.b

26

Table 5 Summary of Returns to Scale by Industry Sector, 1984-88 a

Industry CRS DRS IRS Cases* * *

TeaTobaccoWearing ApparelPrinting and PublishingSoaps, Perfumes and CosmeticsPlastic ProductsFabricated Metal Products

Total (Observations)Percentage (Observations)

4352578

3415.11

197

171023

6

6428.44

57 10 18 8 13 15 6

12756.44

80204020202520

225100.0

Notes:Based on the intertemporal frontier (pooled sample) for each respective sector.a

CRS = Constant Returns to Scale, DRS = Decreasing Returns to Scale, IRS = Increasing*

Returns to Scale

27

Table 6 Descriptive Statistics for Variables in the Panel Regression Model

Variable Description Mean S.D Min Max

Firm Attributes

MS

Industry Attributes

HHI

IMPS

EXPS

GROW

Public Policy

TARF

Other Variables

TEA

TOB

WEA

PRI

SOA

PLA

FAB

= market share of the firm

= the Herfindahl-Hirschman Index.

= import share.

= export shares.

= annual growth rate of domestic

production.

= nominal tariff levels for manufactured

imports.

= 1 if Tea, and zero otherwise,

= 1 if Tobacco

= 1 if Wearing Apparel

= 1 if Printing and Publishing

= 1 if Soaps, Perfumes and Cosmetics

= 1 if Plastic Products

= 1 if Fabricated Metal Products

0.1556

0.2354

0.1760

0.2502

0.3069

0.3946

0.3556

0.0889

0.1778

0.0889

0.0889

0.1111

0.0889

0.2067

0.1489

0.2426

0.2979

0.7564

0.4404

0.4797

0.2852

0.3832

0.2852

0.2852

0.3150

0.2852

0.0048

0.0647

0.0027

0.0000

-0.866

0.0070

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.9067

0.7506

0.9581

1.0000

3.4460

2.6180

1.0000

1.0000

1.0000

1.0000

1.0000

1.0000

1.0000

28

Table 7 Sources of Technical Efficiency: Panel Regression Models

Variables Overall Technical Efficiency (OTE) Scale Efficiency (STE)

Window Analysis Intertemporal

Frontier

Window

Analysis

Intertemporal

Frontier

(1) FE (2) RE (3) FE (4) RE (5) FE (6) RE (7) FE (8) RE

Constant

MS(-1)

HHI

IMPS

EXPS

GROW

TARF

TEA

WEA

PRI

PLA

FAB

-

0.2229

(3.31)a

-0.5131

(-1.69)c

0.3654

(1.37)

-0.0810

(-0.48)

-0.0482

(-0.86)

0.0174

(0.23)

-0.3256

(-2.36)b

0.2060

(1.20)

-0.1804

(-1.34)

-0.3321

(-2.38)b

-0.1245

(-1.46)

0.8252

(7.64)a

0.1467

(1.91)c

-0.3563

(-1.66)c

0.2740

(2.10)b

0.1956

(1.82)c

0.0104

(0.45)

-0.4833

(-1.22)

-0.2883

(-4.35)a

-0.0241

(-0.29)

-0.2126

(-2.59)a

-0.1087

(-0.97)

0.1236

(1.54)

-

0.2995

(3.82)a

-0.3754

(-1.17)

0.1031

(0.34)

0.1609

(0.96)

0.0413

(0.71)

-0.1178

(-1.67)c

-0.4964

(-3.19)a

0.2676

(1.38)

-0.2626

(-1.81)c

-0.0898

(-0.57)

0.0078

(0.07)

0.7569

(6.61)a

0.2235

(3.25)a

-0.1649

(-0.73)

0.0549

(0.17)

0.0186

(0.113)

0.0076

(0.29)

-0.1041

(-2.36)b

-0.3349

(4.61)a

0.0019

(0.02)

-0.4154

(-5.16)a

0.0256

(0.22)

0.1174

(1.35)

-

0.1412

(2.57)b

-0.462

(-2.00)b

0.1293

(0.61)

-0.1230

(-1.12)

-0.0807

(-1.85)c

0.0565

(1.00)

-0.4339

(-5.33)a

0.2742

(1.95)c

-0.0990

(-0.96)

-0.1831

(-1.68)c

-0.1119

(-1.70)c

0.9815

(9.97)a

0.0705

(1.16)

-0.3537

(-1.82)c

0.1353

(1.08)

0.0154

(0.16)

-0.0277

(-1.26)

0.0073

(0.19)

-0.1637

(-2.64)a

-0.0535

(-0.69)

-0.2168

(-3.09)a

-0.0936

(-0.92)

0.0557

(0.75)

-

0.1790

(2.87)a

-0.2983

(-1.40)

-0.2613

(-1.26)

0.0787

(0.71)

-0.0015

(-0.04)

0.0181

(0.40)

-0.4574

(-4.42)a

0.3800

(2.68)a

-0.1330

(-1.28)

0.0575

(1.51)

0.1022

(1.70)c

0.8482

(8.24)a

0.0988

(1.55)

-0.0401

(-0.20)

-0.0990

(-0.76)

0.0642

(0.63)

0.0033

(0.14)

-0.0325

(-0.83)

-0.1242

(-1.91)c

0.0540

(0.67)

-0.3176

(-4.33)a

0.1235

(1.17)

0.1570

(2.02)b

F

Hausman

se

Nobs

0.5025

1.1594

-

0.1973

180

0.3505

-

51.518a

0.2006

180

0.5940

0.8840

-

0.2041

180

0.4964

-

7.9353

0.2023

180

0.4555

1.9138a

-

0.1550

180

0.1686

-

59.514a

0.1707

180

0.5338

1.8987a

-

0.1624

180

0.2901

-

114.89a

0.1786

180

Notes: MS(-1) is the lagged market share, and the excluded industry dummies are TOB and SOA. The figures in

parenthesis are t-statistics (based on heteroskedastic-consistent standard errors for the fixed effects model).

The superscripts a, b and c are levels of significance at 1%, 5% and 10%, respectively. FE = fixed effects

model and RE = random effects model. F is the F-test of A,B=Ai,A and Hausman is the Hausman

specification test.